Lateralized occipito-temporal N1 responses to images of salient distorted finger postures

For humans as social beings, other people’s hands are highly visually conspicuous. Exceptionally striking are hands in other than natural configuration which have been found to elicit distinct brain activation. Here we studied response strength and lateralization of this activation using event-related potentials (ERPs), in particular, occipito-temporal N1 responses as correlates of activation in extrastriate body area. Participants viewed computer-generated images of hands, half of them showing distorted fingers, the other half showing natural fingers. As control stimuli of similar geometric complexity, images of chairs were shown, half of them with distorted legs, half with standard legs. The contrast of interest was between distorted and natural/standard stimuli. For hands, stronger N1 responses were observed for distorted (vs natural) stimuli from 170 ms post stimulus. Such stronger N1 responses were found for distorted hands and absent for distorted chairs, therefore likely unrelated to visuospatial processing of the unusual distorted shapes. Rather, N1 modulation over both hemispheres – but robustly right-lateralized – could reflect distorted hands as emotionally laden stimuli. The results are in line with privileged visual processing of hands as highly salient body parts, with distortions engaging neural resources that are especially activated for biological stimuli in social perception.


Supplementary Methods
Hand stimuli were created using realistic 3D models of actual people's hands who participated in previous experiments conducted by the authors. In order to create these models, right hand dorsal and palmar sides were photographed (approximately 30-40 pictures each) from multiple angles, and then the pictures were uploaded separately onto Autodesk® 123D Catch. This free software allows photos to be uploaded, then through intensity mapping algorithms, creates a 3D mesh of the uploaded photos for each side of the hand. The models created through this software were then processed using Blender 3D®, a free 3D modelling software (https://www.blender.org). for 3D mesh post-processing and texture fixes.
Blender provides an API which allows easy access to edit any interfaced data and allowed applying distortions to the hands and chair stimuli consistently. For each of the hand models, the dorsal side with the wrist structure was parallel to the camera. Then the structure controlling the wrist was rotated by 1. 5°, 1°, 8.3° (x,y,z) in order to mimic a similar perspective to that of viewing your own hand. A 40° rotation was applied anticlockwise along the x-and z-axis for each finger (except the thumb) to the proximal interphalangeal joint (between the proximal and intermediate phalanges) in order to mimic an abnormality as they might occur after an accident (see Figure 1A). Saturation was set to 0, contrast to 2.5, and brightness to 0.8, Then hands were rendered with Blender camera presets (12:11 aspect ratio) with a focal length 35 mm, and 800 x 800 pixels resolution. Using this procedure, 6 individuals' hands were modelled. Due to inter-individual differences in hand size and shape, manual adjustments were made in order to keep the ratio of hand to camera as similar between models as possible. Per hand, 8 images (2 per finger) of distorted finger postures were created. Out of these, 4 images (1 per finger) were randomly selected as stimuli for the experiment. For natural finger postures, 4 images of each hand were created. The whole procedure resulted in 48 images of hands.
Chair stimuli were created using a similar method through modification of 6 freely available templates (http://www.blendswap.com/blends/view/40140, user sizzler, license: CC-BY). Using Blender 3D, distortions were applied to the legs of the chair, in order to create a geometrically matched control for distorted hands As the legs of the chairs do not have any specific landmarks, the distortions were applied at 1/3 third of the leg proximally to the seating base (see Figure 1A) in order to mimic the placement of the distortions on the fingers.
The chairs were placed in similar position as the hands, with legs -as control proxies for the fingers -pointing upwards in diagonal fashion, and the chair was rotated 45° in the z-axis to maximize viewing of all the legs.
Using this procedure, 6 different chairs were modelled. The same script (including camera parameters) was used to create the distortions for each of the chairs, creating 8 distorted images (2 distortions for each of 4 legs). Out of these, 4 images (1 per leg) were randomly selected as stimuli for the experiment. For standard chairs, 4 images of each chair were created. The whole procedure resulted in 48 images of hands.
In order to hide the end of the hand model, as this may distract from the finger distortion, a Gaussian filter was applied to centre of the image such that the edges of the images, and consequently the end of model, were blurred while the centre of the image remained visible (see Figure 1A). Supplementary Table 1a: P1 amplitudes, separate by stimulus type, configuration, and hemisphere (electrodes PO3 and PO4). The reasons to choose PO3 and PO4 for exploratory analysis of P1 responses were (a) clearly defined P1 responses in grand averages and (b) lowest variability across subjects (measured as SD, compared with P7/P8, P9/10, and PO7/PO8, the 6 electrodes of interest for N1 in main analysis). PO3/PO4 were also among the electrodes studied in Thierry et al. (2007).

Supplementary Results and Discussion
Complementary to the hypothesis-driven analysis of N1 responses (with electrodes of interest and time window chosen on the basis of research literature), exploratory analysis was performed on responses to hands later than N1. The purpose of this analysis was to compare EEG responses in the current study with the bilateral MEG responses to distorted hands in an earlier MEG study (Avikainen et al. 2003). In the MEG study, differences between distorted and natural hands started at 260 ms after stimulus onset and were most consistent across subjects in a 400 to 600 ms time window. Consequently, the time window for EEG analysis was chosen to start at 250 ms and to end at 500 ms (end of segmented EEG trials).
In ANOVA of EEG amplitudes (averaged between 250 and 500 ms), neither of the main effects (configuration: distorted, natural; hemisphere: left, right) nor the configuration X hemisphere interaction was significant (Supplementary Table 4a).
The MEG parameter was source strength (always positive) whereas the EEG parameter was amplitude relative to baseline (positive or negative). Therefore, in an additional analysis step, EEG responses were assessed in terms of root-mean-square (rms) amplitudes (always positive). Again, in ANOVA of rms EEG amplitudes (250 to 500 ms), neither of the main effects (configuration; hemisphere) nor the configuration X hemisphere interaction was significant (Supplementary Table 4b).
Although Critically, the MEG study and our current study were also different in terms of the task for the subject: in the two conditions of the MEG study, subjects either detected stimulus repeats in a 1-back task or imitated the previously seen hand posture when an imperative (non-hand) stimulus was shown. In the current EEG study, subjects had to mentally count occurrences of a shadow superimposed on some of the stimuli. The MEG study therefore required processing of hand postures for demands of the 1-back and imitation tasks whereas in our study postures were entirely irrelevant to the task. It is all the more remarkable that our hand postures, although irrelevant to the task, still elicited distorted vs natural response differences at N1 latency (see Discussion, section "Task demands").